This post is more than 5 years old
2 Intern
•
5.7K Posts
0
3515
Dropped packets in portchannel without bandwidth issue
Hello experts,
I'm seeing the following: 2 Cisco MDS switches connected over a portchannel which consists of 2 ISLs. Each ISL is running at 4Gb, so the portchannel is 8Gb. We're seeing lots of dropped packets during 11PM and 6AM and II'd like to know why this happens. The physical layer seems ok, because during the day we don't see any errors on the ports at all.
Somebody says it could be because some storage cannot process data fast enough and the data flow has to slow down because of that, but in that case I expect to see dropped packets on the ports the storage is connected to, not some ISLs!
Does anyone know why this happens?
healyj
141 Posts
1
December 12th, 2017 07:00
Hi there,
In our efforts to clean up the forum, we came across your question / statement.
If the question / statement is still valid, not expired and you need an update please reach out again and we try to get it answered.
As for now we will set it to “answered.”
Regards,
Jim
RRR
2 Intern
2 Intern
•
5.7K Posts
0
June 21st, 2012 02:00
Our monitoring tool reports that about half of the bandwidth is available during these hours.
dynamox
2 Intern
2 Intern
•
20.4K Posts
0
June 21st, 2012 04:00
can you remove one of the ISL links from this portchannel and see what happens? Add this ISL back and then remove the other one .
RRR
2 Intern
2 Intern
•
5.7K Posts
0
June 22nd, 2012 04:00
Thanks, I will try that.
There aren't any CRC errors btw, only dropped packets. The SR we had open mentioned that this could happen when large loads of traffic were going over the portchannel and the storage couldn't handle that. The portchannel is 2 x 4Gb and the peek load was only 1.2Gbps, so I doubt if that was the problem. Any storage system we have out there can easily handle 150MBps. I'll advise to do the ISL trick and let you know.
dynamox
2 Intern
2 Intern
•
20.4K Posts
0
July 12th, 2012 07:00
are you staying with Cisco or going to Brocade ?
RRR
2 Intern
2 Intern
•
5.7K Posts
0
July 12th, 2012 07:00
My employer stays with Cisco and as long as you have knowledgable people managing it, that's fine. I'm stationed at a customer's site and over here we have both. A transition from Cisco to Brocade is going on, but that might just be politics. IMHO it's a good thing, because I've always loved Brocade, but we're now reaching certain limits there as well. Limits that I never had to worry about a few years ago. Things like fillword 3's, 8Gb, attenuation, FCIP over MPLS, DWDM... it's all very exciting
and Brocade sure has it's chalenges there as well. But we'll conquer them!
RRR
2 Intern
2 Intern
•
5.7K Posts
0
July 12th, 2012 07:00
Haven't had an update from the guys of operations yet. They did however found another issue with some sort of memory leak in one of the 9222is we have, so we're going to upgrade the SAN as soon as possible to get rid of that one first and hopefully the dropped packets are gone then as well
felipon1
96 Posts
1
July 13th, 2012 01:00
Hi,
Do you see frame discards in any host port? I have seen a similar behavior when a host cannot process the incoming/outgoing load and frames start to queue in the F-port. As buffers get filled up, frames are start to queue in the ISLs buffer and will end up being discarded.
Increasing the number of B2B credits configured can help avoiding this behavior, but usually Cisco is very generous in this case...
RRR
2 Intern
2 Intern
•
5.7K Posts
0
July 13th, 2012 04:00
Another good tip! The thing is however that the 2 switches connected here are sitting next to eachoter (well, above each other in the same rack). So I would think that B2B credits cannot be the issue here. Right?